Watch Tiberiu Socaci build a custom web interface with Claude AI, Make.com webhooks and Airtable, then connect user actions to automation and an AI agent.
Recorded on 1 July 2026, this webinar shows how Tiberiu Socaci uses Claude AI to create a tailored interface and Make.com to serve the page, load data and process user actions. The recording includes working examples and a live build.
How the interface works
A visitor opens a URL backed by a Make webhook. A Make scenario returns an HTML and JavaScript page, together with any data the interface needs. When someone submits a form, approves an item or sends a message, the page calls another webhook that runs the next scenario. Tiberiu shows where each request appears in Make's execution history, so the team can inspect the flow and change it.
Claude helps write the front end, including its layout and behaviour. Make handles connections to systems such as Airtable and the actions that follow. Tiberiu presents this as an option for focused internal or customer-facing workflows, especially when a small interface does not need a separate application server.
Examples shown in the webinar
- Lead intake and call routing. A multi-step form enriches a company with Apollo and shows the appropriate team member's booking calendar.
- Marketing content review. A team member chooses an idea, asks an AI agent to draft copy and sends the result back as a draft for review.
- Specialist booking. A customer sees availability from Airtable, selects a person and time, then submits a booking through a Make scenario.
- Bus maintenance. A worker uploads a bus photo, identifies the vehicle and asks an agent about its repair history.
The live build
Tiberiu builds a product-review page with Claude and a Make webhook response. The page receives product records from Airtable, collects approval decisions and feedback, and sends that feedback to another scenario. He then connects a product-level chat to an AI agent and tests two ways of supplying context: passing the product details with each message or letting the agent retrieve a record by ID when needed.
The demonstration also shows the limits of a quick build: sample data and webhook URLs need to be replaced, the response format needs checking, and each action needs to be tested before the interface is ready for real users.
